altos/arm: Align data so that gcc 5.4 doesn't do byte-accesses. Add -Wcast-align

Gcc 5.4.1 tracks alignment of data through assignments, so that a
uint32_t pointer which comes from byte-aligned uint8_t data:

extern uint8_t foo[];

	uint32_t	*q = (void *) foo;

Fetches and stores through this pointer are done bytewise. This is
slow (meh), but if q references a device register, things to bad very
quickly.

This patch works around this bug in the compiler by adding
__attribute__((aligned(4))) tags to some variables, or changing them
from uint8_t to uint32_t. Places doing this will now be caught as I've
added -Wcast-align to the compiler flags. That required adding (void
*) casts, after the relevant code was checked to make sure the
compiler could tell that the addresses were aligned.

Signed-off-by: Keith Packard <keithp@keithp.com>

5 files changed, 7 insertions, 12 deletions

diff --git a/src/stm/Makefile.defs b/src/stm/Makefile.defs
index c3d2707f..0ba86f5a 100644
--- a/src/stm/Makefile.defs
+++ b/src/stm/Makefile.defs
@@ -27,7 +27,7 @@ LIBS=$(PDCLIB_LIBS_M3) -lgcc
 WARN_FLAGS=-Wall -Wextra -Werror
 
 AO_CFLAGS=-I. -I../stm -I../kernel -I../drivers -I../math -I.. $(PDCLIB_INCLUDES)
-STM_CFLAGS=-std=gnu99 -mlittle-endian -mcpu=cortex-m3 -mthumb \
+STM_CFLAGS=-std=gnu99 -mlittle-endian -mcpu=cortex-m3 -mthumb -Wcast-align \
 	-ffreestanding -nostdlib $(AO_CFLAGS) $(WARN_FLAGS)
 
 LDFLAGS=-L../stm -Wl,-Taltos.ld
diff --git a/src/stm/ao_arch_funcs.h b/src/stm/ao_arch_funcs.h
index 18ca20da..a9d0fa34 100644
--- a/src/stm/ao_arch_funcs.h
+++ b/src/stm/ao_arch_funcs.h
@@ -375,7 +375,7 @@ ao_arch_irq_check(void) {
 static inline void
 ao_arch_init_stack(struct ao_task *task, void *start)
 {
-	uint32_t	*sp = (uint32_t *) (task->stack + AO_STACK_SIZE);
+	uint32_t	*sp = (uint32_t *) ((void*) task->stack + AO_STACK_SIZE);
 	uint32_t	a = (uint32_t) start;
 	int		i;
 
@@ -413,16 +413,11 @@ static inline void ao_arch_save_stack(void) {
 	uint32_t	*sp;
 	asm("mov %0,sp" : "=&r" (sp) );
 	ao_cur_task->sp = (sp);
-	if ((uint8_t *) sp < &ao_cur_task->stack[0])
-		ao_panic (AO_PANIC_STACK);
 }
 
 static inline void ao_arch_restore_stack(void) {
-	uint32_t	sp;
-	sp = (uint32_t) ao_cur_task->sp;
-
 	/* Switch stacks */
-	asm("mov sp, %0" : : "r" (sp) );
+	asm("mov sp, %0" : : "r" (ao_cur_task->sp) );
 
 	/* Restore PRIMASK */
 	asm("pop {r0}");
diff --git a/src/stm/ao_eeprom_stm.c b/src/stm/ao_eeprom_stm.c
index 05f880b8..4f477122 100644
--- a/src/stm/ao_eeprom_stm.c
+++ b/src/stm/ao_eeprom_stm.c
@@ -83,7 +83,7 @@ ao_intflash_write32(uint16_t pos, uint32_t w)
 {
 	volatile uint32_t	*addr;
 
-	addr = (uint32_t *) (stm_eeprom + pos);
+	addr = (uint32_t *) (void *) (stm_eeprom + pos);
 
 	/* Write a word to a valid address in the data EEPROM */
 	*addr = w;
@@ -96,7 +96,7 @@ ao_intflash_write8(uint16_t pos, uint8_t d)
 	uint32_t	w, *addr, mask;
 	uint8_t		shift;
 	
-	addr = (uint32_t *) (stm_eeprom + (pos & ~3));
+	addr = (uint32_t *) (void *) (stm_eeprom + (pos & ~3));
 
 	/* Compute word to be written */
 	shift = (pos & 3) << 3;
diff --git a/src/stm/ao_usb_stm.c b/src/stm/ao_usb_stm.c
index f2b8ea94..9d72844e 100644
--- a/src/stm/ao_usb_stm.c
+++ b/src/stm/ao_usb_stm.c
@@ -139,7 +139,7 @@ static inline uint32_t set_toggle(uint32_t 	current_value,
 
 static inline uint32_t *ao_usb_packet_buffer_addr(uint16_t sram_addr)
 {
-	return (uint32_t *) (stm_usb_sram + 2 * sram_addr);
+	return (uint32_t *) (((void *) ((uint8_t *) stm_usb_sram + 2 * sram_addr)));
 }
 
 static inline uint32_t ao_usb_epr_stat_rx(uint32_t epr) {
diff --git a/src/stm/stm32l.h b/src/stm/stm32l.h
index 463125e2..be1e1d65 100644
--- a/src/stm/stm32l.h
+++ b/src/stm/stm32l.h
@@ -1961,7 +1961,7 @@ union stm_usb_bdt {
 
 #define STM_USB_BDT_SIZE	8
 
-extern uint8_t stm_usb_sram[];
+extern uint8_t stm_usb_sram[] __attribute__ ((aligned(4)));
 
 struct stm_exti {
 	vuint32_t	imr;